A century of warming on Caribbean reefs
Colleen B. Bove1,2*, Laura Mudge1,3, and John F. Bruno1
1 The Department of Biology, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, 27599-3280 USA
2 The Department of Biology, Boston University, Boston, Massachusetts, 02215 USA
3 Integral Consulting Inc., Annapolis, Maryland 21401 USA
*Corresponding author: colleenbove@gmail.com
Abstract:
The world’s oceans are warming at an unprecedented rate, causing dramatic changes to coastal marine systems, especially coral reefs. We used three complementary ocean temperature databases (HadISST, Pathfinder, and OISST) to quantify change in thermal characteristics of Caribbean coral reefs over the last 150 years (1871–2020). These sea surface temperature (SST) databases included in situ and satellite-derived measurements at multiple spatial resolutions. We also compiled a Caribbean coral reef database identifying 5,326 unique reefs across the region. We found that Caribbean reefs have been warming for at least a century. Regionally reef warming began in 1915, and for four of the eight Caribbean ecoregions we assessed, significant warming was detected for the latter half of the nineteenth century. Following the global mid-twentieth century stasis, warming resumed on Caribbean reefs in the early 1980s in some ecoregions and in the 1990s for others. On average, Caribbean reefs warmed by 0.18 °C per decade during this period, ranging from 0.17 °C per decade on Bahamian reefs (since 1988) to 0.26 °C per decade on reefs within the Southern and Eastern Caribbean ecoregions (since 1981 and 1984, respectively). If this linear rate of warming continues, these already threatened ecosystems would warm by an additional ~1.5°C on average by 2100. We also found that marine heatwave (MHW) events are increasing in both frequency and duration across the Caribbean. Caribbean coral reefs now experience on average 5 MHW events annually, compared to 1 per year in the early 1980s, with recent events lasting on average 14 days. These changes in the thermal environment, in addition to other stressors including fishing and pollution, have caused a dramatic shift in the composition and functioning of Caribbean coral reef ecosystems.
Citation:
Repository DOI hosted on Zenodo 
Caribbean Ecoregion Coral Reefs
Figure 1
FALSE OGR data source with driver: ESRI Shapefile
FALSE Source: "/Users/colleen/Dropbox/Git/CaribbeanSST/data/EcoRegions", layer: "Caribbean_ecoregions"
FALSE with 17 features
FALSE It has 9 fields
Fig 1. Caribbean coral reef site locations and ecoregion designation. The colour of each reef represents the designated ecoregion and n denotes the number of unique reef locations within that ecoregion. Map layer was acquired from Natural Earth (https://www.naturalearthdata.com).
SST Extraction and Analysis
Figure 2
Fig 2. Historic SST trend on Caribbean coral reefs (1871–2020). Long-term SST records (HadISST) on Caribbean coral reefs depicting A) mean monthly SST each year (represented by line colour: blue to red) and B) annual mean SST with significant warming based on the first derivative of the GAM slope (1915 and 1994 highlighted in the yellow boxes). The GAM smoothed annual mean SST over time is represented by the black line with significantly warming periods (red) identified over the curve. The annual maximum (red line) and annual minimum (blue line) SST are also depicted, along with the overall mean SST for all sites over the entire period (27.3 °C; grey dashed line). The grey ribbon represents the 95% confidence interval around the annual SST mean through time. Vertical grey lines represent year of significant warming across ecoregions.
Figure 3
Fig 3. Historic SST trends on coral reefs within ecoregions (1871–2020). Long-term SST records (HadISST) on Caribbean coral reefs separated by ecoregion depicting the GAM smoothed annual mean SST through time (black line) with significantly warming periods (red line and yellow background; see S3 Table) identified over the curve. The grey dashed horizontal line denotes the mean SST over the entire period, the red line depicts the annual maximum SST, the blue line depicts the annual minimum SST, and the grey ribbon represents the 95% confidence interval around the true annual SST mean for each ecoregion.
Figure 6
Fig 6. Climate stripe diagram depicting the mean annual temperature recorded on Caribbean coral reefs based on the HadISST database. Warmer temperatures are depicted in red (maximum annual SST of 28.0) and cooler annual temperatures are in blue (minimum annual SST of 26.6) (figure inspired by Professor Ed Hawkins Show Your Stripes).
Figure S1
Supplemental Figure 1. Historic SST records (1981-2019; Pathfinder) on Caribbean coral reefs depicting A) mean monthly SST each year (represented by line colour: blue to red) and B) GAM smoothed annual mean SST time (black line), annual maximum (red line), and annual minimum (blue line) SST. The grey dashed horizontal line denotes the overall mean SST for all sites over the entire period (27.23 °C) and the grey ribbon represents the 95% confidence interval around the true annual SST mean through time.
Figure S2
Supplemental Figure 2. Comparison of monthly SST (°C) recorded in the HadISST and Pathfinder databases from 1982-2019. SST values were regressed against one another for the same time for comparison of the relationship between the databases. Points are depicted by month (cooler months in blue and warmer months in red) and the black line represents the resulting linear regression model (R2 = 0.97; P < 0.001).
Figure S3
Supplemental Figure 3. Comparison of HadISST (1871-2020) and Pathfinder (1981-2019) SST recorded on Caribbean coral reef locations. The high-resolution Pathfinder is represented as darker data over the long-term HadISST. Both datasets are represented by GAM smoothed annual mean SST time (solid line), annual maximum (red line), and annual minimum (blue line) SST. The dashed horizontal line denotes the overall mean SST for all sites over the entire period and the grey ribbon represents the 95% confidence interval around the true annual SST mean through time.
Figure S4
Supplemental Figure 4. Historic SST trends on coral reefs within Caribbean ecoregions (1871 - 2020) with corresponding reef locations (see Figures 1, 3 in the main text). The colour of each reef location and box around long-term SST (HadISST) plots represent the designated ecoregion. Plots depict SST data with GAM smoothed annual mean SST time (black line), annual maximum (red line), and annual minimum (blue line) SST. The grey dashed horizontal line denotes the mean SST over the entire period and the grey ribbon represents the 95% confidence interval around the true annual SST mean for the A) Bahamian, B) Eastern Caribbean, C) Floridian, D) Greater Antilles, E) Gulf of Mexico, F) Southern Caribbean, G) Southwestern Caribbean, and H) Western Caribbean ecoregions.
HadISST GIF
Long-term SST records (HadISST) on Caribbean coral reefs depicting mean monthly SST each year (represented by line colour: blue to red).
Caribbean Basin Maps and Marine Heatwaves
Figure 4
All Marine Heatwave analyses were performed by Laura Mudge (GitHub: Lmudge13)
Laura’s code/analyses are included in this script and identified within code chunks when pulled or modified from her original work.
Fig 4. Warming patterns throughout the Caribbean Sea. Increasing warming events across the Caribbean depicted through A) rate of SST change (°C per decade) from 1981 to 2019 (Pathfinder; mean slope 0.17 °C per decade) and B) increasing marine heatwave events (slope of counts per year). Grey ocean area was not included in these analyses. Map layer was acquired from Natural Earth (https://www.naturalearthdata.com).
Figure 5
Fig 5. MHW trends (1981–2018) across Caribbean coral reefs. Temperature data are based on OISST gridded data to determine A) marine heat wave (MHW) frequency (number events per year) with Nagelkerke pseudo R2; B) MHW duration (number days per event) with linear model R2 ; and C) return time (number days per event) since the previous MHW event with linear model R2 reported. Points denote annual mean values (±SD) and blue lines represent linear (lm or glm) trends.
Figure S5
Supplemental Figure 5. Significance of A) rate of SST change (°C per decade) and B) number of marine heatwave events per year across the Caribbean depicted in Figure 4. Grey ocean area was not included in these analyses.
Figure S6
Supplemental Figure 6. A) Rate of SST change (°C per decade) over the duration of the HadISST database across the Caribbean from 1871 to 2020 (mean slope 0.04 ± 0.014 °C per decade) and B) significance of rate of SST change. Grey ocean area was not included in these analyses.
Figure S7
Supplemental Figure 7. A) Rate of SST change (°C per decade) across the Caribbean from 1981 to 2020 (HadISST; mean slope 0.17 ± 0.046 °C per decade) and B) significance of rate of SST change. Grey ocean area was not included in these analyses.
Figure S8
Supplemental Figure 8. MHW trends (1981 - 2018) across Caribbean coral reefs by ecoregion. Temperature data are based on OISST gridded data to determine frequency (number events per year), duration (number days per event), return time (number days per event) since the previous event, onset rate (°C per day) from start until peak intensity, peak intensity (°C), and total days reefs experience MHWs per year. Points denote annual mean values (±SD) and blue lines represent linear (lm or glm) trends within each ecoregion (see Figures 1, Supplemental Figure 4 for ecoregion locations). Frequency, duration, and return time across all Caribbean coral reefs are depicted in Figure 5 in the main text.
Tables
Table 1
Table 1. Estimated warming rates from both HadISST and Pathfinder databases for different temporal ranges. Coral reef values are means of all 5,326 reef locations included in the study. The year 1994 was estimated as the beginning of the most recent period of warming across all Caribbean coral reefs (see S1 Fig). The HadISST data can be accessed at www.metoffice.gov.uk/hadobs/hadisst/, the Pathfinder can be accessed at www.ncei.noaa.gov/products/avhrr-pathfinder-sst, and the OISST can be accessed at www.ncei.noaa.gov/products/optimum-interpolation-sst.
| Temperature parameter | HadISST (1871-2020) | HadISST (1981-2020) | HadISST (1994-2020) | Pathfinder (1981-2019) | Pathfinder (1994-2019) | OISST (1982-2019 |
|---|---|---|---|---|---|---|
| Caribbean Basin (°C per decade) | 0.04 | 0.17 | 0.2 | 0.17 | NA | NA |
| Caribbean Basin (total °C for period) | 0.6 | 0.68 | 0.54 | 0.66 | NA | NA |
| Caribbean Reefs (°C per decade) | 0.04 | 0.15 | 0.17 | 0.19 | 0.18 | NA |
| Caribbean Reefs (total °C for period) | 0.6 | 0.6 | 0.46 | 0.74 | 0.47 | NA |
| Caribbean Basin (increasing frequency of MHW per year) | NA | NA | NA | NA | NA | 0.05 |
Table 2
Table 2. Published reports of global ocean surface warming rates.
| Study | °C per decade | Years |
|---|---|---|
| Casey and Cornillon 2001 41 | 0.14 | 1960–1990 |
| Lawrence et al. 2004 87 | 0.09–0.13 | 1985–2000 |
| Good et al. 2007 88 | 0.17 | 1985–2004 |
| Burrows et al. 2011 89 | 0.07 | 1960–2009 |
| USGCRP 2017 90 | 0.15 | 1900–2016 |
Table S1
Table S1. Marine heatwave (MHW) properties examined in this study, as developed by Hobday et al. 2016.
| Metric | Units | Description |
|---|---|---|
| Frequency | Number of events | Number of discrete MHW events |
| Total MHW days | days | The total number of days a location experienced a MHW per year |
| Duration | days | The number of days between the start and end date of each distinct MHW event |
| Peak intensity | °C | The maximum temperature, above the seasonal varying climatological mean, reached during the MHW event |
| Onset rate | °C per day | The rate of change in temperature between the event start date and date of peak intensity |
| Return time | days | The number of days elapsed since a previous MHW event in that location |
Table S2
Supplemental Table 2. Description of the SST or MHW metrics assessed in this manuscript with details on the source of the temperature data (database), the scale of the measure, any data transformations completed for statistical analyses (transformation), the statistical test performed, and the R package used for each test. Further details on these methods can be found on the corresponding GitHub repository github.com/seabove7/CaribbeanSST and on Zenodo (DOI: 10.5281/zenodo.4751658).
| Metric | Database | Scale | Transformation | Statistical test | Package (version) |
|---|---|---|---|---|---|
| Rate of SST change | HadISST | basin | NA | GLS; correlation structure of time for autocorrelation | nlme (3.1.151)2 |
| Rate of SST change | HadISST | reefs | NA | GAMM; cubic spline for year and cyclic cubic regression spline for month and correlation structure of time for autocorrelation | mgcv (1.8.36)3 |
| Rate of SST change | Pathfinder | basin | NA | GLS; correlation structure of time for autocorrelation | nlme (3.1.151)2 |
| MHW frequency | OISST | reefs | NA | GLM; poisson with log link | stats (3.6.3)4 |
| MHW duration | OISST | reefs | log | OLS model | stats (3.6.3)4 |
| MHW return time | OISST | reefs | log | OLS model | stats (3.6.3)4 |
| MHW total # events | OISST | reefs by ecoregion | NA OLS model | lme4 (1.1.27.1)5 | |
| MHW duration | OISST | reefs by ecoregion | log | OLS model | lme4 (1.1.27.1)5 |
| MHW return time | OISST | reefs by ecoregion | log | OLS model | lme4 (1.1.27.1)5 |
| MHW max intensity | OISST | reefs by ecoregion | log | OLS model | lme4 (1.1.27.1)5 |
| MHW onset | OISST | reefs by ecoregion | log | OLS model | lme4 (1.1.27.1)5 |
| MHW frequency | OISST | reefs by ecoregion | NA | GLM; poisson with log link | lme4 (1.1.27.1)5 |
Table S3
Supplemental Table 3. Mean ocean warming rate (°C per decade; with 95% confidence interval) and total increase in temperature (°C) on coral reefs within each Caribbean ecoregion since the noted year of inflection point. The inflection point for each ecoregion was identified as the year in which annual warming rates significantly increased based on the first derivative of the GAM curve for each ecoregion (see Figures 3 and S5).
| Ecoregion | Inflection point | Mean rate (°C per decade) | Lower confidence interval | Upper confidence interval | Warming extent (°C) |
|---|---|---|---|---|---|
| Bahamian | 1988 | 0.17 | 0.161 | 0.182 | 0.53 |
| Eastern Caribbean | 1984 | 0.26 | 0.260 | 0.267 | 0.91 |
| Floridian | 1993 | 0.22 | 0.207 | 0.231 | 0.57 |
| Greater Antilles | 1986 | 0.16 | 0.160 | 0.168 | 0.53 |
| Gulf of Mexico | 1981 | 0.21 | 0.186 | 0.241 | 0.80 |
| Southern Caribbean | 1981 | 0.26 | 0.258 | 0.268 | 0.99 |
| Southwestern Caribbean | 1982 | 0.20 | 0.194 | 0.204 | 0.74 |
| Western Caribbean | 1999 | 0.24 | 0.235 | 0.253 | 0.48 |
Table S4
Table S4. Model results for MHW trends on coral reefs in the Caribbean basin between 1981-2018. Frequency was modelled using a glm model with poisson distribution and log link. Estimates, standard error, z scores (statistic), and p-values are reported. Nagelkerke pseudo r-squared used to assess goodness of fit. MHW duration and return time were log transformed before modelling with ols models. Estimate, standard error, t value (statistic) and p-value are reported along with multiple and adjusted r-squared.
| Estimate | Standard error | Statistic | P-value | |
|---|---|---|---|---|
| Frequency | ||||
| (Intercept) | -93.154 | 0.892 | -104.406 | < 0.001 |
| year | 0.047 | 0.000 | 105.559 | < 0.001 |
| R2 | 0.480 | |||
| Duration | ||||
| (Intercept) | -36.508 | 0.666 | -54.790 | < 0.001 |
| year | 0.019 | 0.000 | 58.157 | < 0.001 |
| R2 | 0.185 | |||
| Return Time | ||||
| (Intercept) | 87.968 | 1.668 | 52.750 | < 0.001 |
| year | -0.041 | 0.001 | -49.840 | < 0.001 |
| R2 | 0.148 | |||
Table S4
Table S5. Mean decadal value of MHW parameters (frequency, duration, and return time) for the entire basin and by ecoregion.
| Region | 1980 | 1990 | 2000 | 2010 |
|---|---|---|---|---|
| Duration (days) | ||||
| Caribbean | 7.6 | 8.3 | 10.8 | 14.5 |
| Bahamian | 8.6 | 8.6 | 10.3 | 12.8 |
| Eastern Caribbean | 6.9 | 8.1 | 15.0 | 19.0 |
| Floridian | 8.9 | 9.5 | 9.6 | 11.3 |
| Greater Antilles | 7.3 | 7.9 | 11.1 | 14.4 |
| Gulf of Mexico | 8.2 | 10.8 | 9.5 | 14.6 |
| Southern Caribbean | 7.5 | 7.5 | 10.3 | 15.7 |
| Southwestern Caribbean | 7.3 | 7.5 | 9.6 | 15.3 |
| Western Caribbean | 7.1 | 8.4 | 8.9 | 11.4 |
| Frequency (events per year) | ||||
| Caribbean | 1.2 | 1.6 | 2.8 | 4.3 |
| Bahamian | 1.6 | 2.0 | 2.6 | 4.2 |
| Eastern Caribbean | 0.5 | 1.0 | 2.7 | 4.1 |
| Floridian | 1.2 | 2.7 | 2.5 | 5.4 |
| Greater Antilles | 1.2 | 1.5 | 2.9 | 4.4 |
| Gulf of Mexico | 1.5 | 2.0 | 3.0 | 4.1 |
| Southern Caribbean | 1.0 | 1.2 | 2.9 | 4.5 |
| Southwestern Caribbean | 1.2 | 1.4 | 3.1 | 4.7 |
| Western Caribbean | 1.6 | 2.0 | 2.7 | 3.2 |
| Return Time (days) | ||||
| Caribbean | 376.7 | 373.2 | 202.4 | 110.8 |
| Bahamian | 304.6 | 264.9 | 201.3 | 122.2 |
| Eastern Caribbean | 721.9 | 651.1 | 236.3 | 94.9 |
| Floridian | 334.2 | 234.7 | 182.0 | 83.1 |
| Greater Antilles | 398.9 | 435.0 | 202.1 | 107.8 |
| Gulf of Mexico | 364.9 | 247.7 | 158.2 | 133.6 |
| Southern Caribbean | 327.8 | 442.4 | 202.6 | 106.4 |
| Southwestern Caribbean | 449.2 | 422.2 | 194.1 | 96.2 |
| Western Caribbean | 299.3 | 257.8 | 204.7 | 145.9 |
Session information
Sea surface temperature and visualization code written by Colleen B. Bove and marine heatwave code written by Laura Mudge.
Session information from the latest compile on 2022-02-10:
## R version 3.6.3 (2020-02-29)
## Platform: x86_64-apple-darwin15.6.0 (64-bit)
## Running under: macOS Catalina 10.15.7
##
## Matrix products: default
## BLAS: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## attached base packages:
## [1] splines grid stats graphics grDevices utils datasets
## [8] methods base
##
## other attached packages:
## [1] transformr_0.1.3 fmsb_0.7.1 performance_0.7.3
## [4] modelr_0.1.8 broom_0.7.9 sjPlot_2.8.9
## [7] ggpubr_0.4.0 gridExtra_2.3 kableExtra_1.3.4
## [10] plotly_4.9.4.1 foreach_1.5.1 mgcv_1.8-36
## [13] nlme_3.1-151 Hmisc_4.5-0 Formula_1.2-4
## [16] survival_3.2-12 lattice_0.20-44 gganimate_1.0.7
## [19] fields_12.5 spam_2.6-0 dotCall64_1.0-1
## [22] binr_1.1 gapminder_0.3.0 gifski_1.4.3-1
## [25] RColorBrewer_1.1-2 xts_0.12.1 zoo_1.8-9
## [28] rnaturalearth_0.1.0 forcats_0.5.1 stringr_1.4.0
## [31] dplyr_1.0.7 purrr_0.3.4 readr_2.0.1
## [34] tidyr_1.1.3 tibble_3.1.3 tidyverse_1.3.1
## [37] openxlsx_4.2.4 rgdal_1.5-23 maptools_1.1-1
## [40] viridis_0.6.1 viridisLite_0.4.0 raster_3.4-13
## [43] sp_1.4-5 ggrepel_0.9.1 ggplot2_3.3.5
## [46] ncdf4_1.17 sf_1.0-2
##
## loaded via a namespace (and not attached):
## [1] utf8_1.2.2 tidyselect_1.1.1 lme4_1.1-27.1
## [4] htmlwidgets_1.5.3 lpSolve_5.6.15 munsell_0.5.0
## [7] codetools_0.2-18 effectsize_0.4.5 units_0.7-2
## [10] withr_2.4.2 colorspace_2.0-2 highr_0.9
## [13] knitr_1.33 rstudioapi_0.13 wk_0.5.0
## [16] ggsignif_0.6.2 labeling_0.4.2 emmeans_1.6.2-1
## [19] rnaturalearthdata_0.1.0 bit64_4.0.5 farver_2.1.0
## [22] datawizard_0.2.0 coda_0.19-4 vctrs_0.3.8
## [25] generics_0.1.0 TH.data_1.0-10 xfun_0.25
## [28] R6_2.5.1 assertthat_0.2.1 vroom_1.5.4
## [31] scales_1.1.1 multcomp_1.4-17 nnet_7.3-16
## [34] rgeos_0.5-5 gtable_0.3.0 sandwich_3.0-1
## [37] rlang_0.4.11 systemfonts_1.0.2 rstatix_0.7.0
## [40] lazyeval_0.2.2 checkmate_2.0.0 s2_1.0.6
## [43] yaml_2.2.1 abind_1.4-5 crosstalk_1.1.1
## [46] backports_1.2.1 tools_3.6.3 ellipsis_0.3.2
## [49] jquerylib_0.1.4 proxy_0.4-26 Rcpp_1.0.7
## [52] plyr_1.8.6 base64enc_0.1-3 progress_1.2.2
## [55] classInt_0.4-3 prettyunits_1.1.1 rpart_4.1-15
## [58] cowplot_1.1.1 haven_2.4.3 cluster_2.1.2
## [61] fs_1.5.0 magrittr_2.0.1 data.table_1.14.0
## [64] reprex_2.0.1 mvtnorm_1.1-2 sjmisc_2.8.7
## [67] hms_1.1.0 evaluate_0.14 xtable_1.8-4
## [70] rio_0.5.27 sjstats_0.18.1 jpeg_0.1-9
## [73] readxl_1.3.1 ggeffects_1.1.1 compiler_3.6.3
## [76] maps_3.3.0 KernSmooth_2.23-20 crayon_1.4.1
## [79] minqa_1.2.4 htmltools_0.5.2 tzdb_0.1.2
## [82] lubridate_1.7.10 DBI_1.1.1 tweenr_1.0.2
## [85] sjlabelled_1.1.8 dbplyr_2.1.1 MASS_7.3-54
## [88] boot_1.3-28 Matrix_1.3-4 car_3.0-11
## [91] cli_3.0.1 parallel_3.6.3 insight_0.14.3
## [94] pkgconfig_2.0.3 foreign_0.8-75 xml2_1.3.2
## [97] svglite_2.0.0 bslib_0.3.1 webshot_0.5.2
## [100] estimability_1.3 rvest_1.0.1 digest_0.6.27
## [103] parameters_0.14.0 rmarkdown_2.10 cellranger_1.1.0
## [106] htmlTable_2.2.1 curl_4.3.2 nloptr_1.2.2.2
## [109] lifecycle_1.0.0 jsonlite_1.7.2 carData_3.0-4
## [112] fansi_0.5.0 pillar_1.6.2 fastmap_1.1.0
## [115] httr_1.4.2 glue_1.4.2 bayestestR_0.10.5
## [118] zip_2.2.0 png_0.1-7 iterators_1.0.13
## [121] bit_4.0.4 class_7.3-19 stringi_1.7.3
## [124] sass_0.4.0 latticeExtra_0.6-29 e1071_1.7-8